kernel: introduce per process FPU struct (PFPU) for more flexible machine specific fpu handling

introducing the PFPU structue which allows the machine specific
code some flexibility on how to handle the FPU process state.

for example, in the pc and pc64 kernel, the FPsave structure is
arround 512 bytes. with avx512, it could grow up to 2K. instead
of embedding that into the Proc strucutre, it is more effective
to allocate it on first use of the fpu, as most processes do not
use simd or floating point in the first place. also, the FPsave
structure has special 16 byte alignment constraint, which further
favours dynamic allocation.

this gets rid of the memmoves in pc/pc64 kernels for the aligment.

there is also devproc, which is now checking if the fpsave area
is actually valid before reading it, avoiding debuggers to see
garbage data.

the Notsave structure is gone now, as it was not used on any
machine.

3 files changed, 12 insertions, 16 deletions

diff --git a/sys/src/9/zynq/dat.h b/sys/src/9/zynq/dat.h
index 7431ed803..59b60ba19 100644
--- a/sys/src/9/zynq/dat.h
+++ b/sys/src/9/zynq/dat.h
@@ -1,13 +1,13 @@
 typedef struct Conf	Conf;
 typedef struct Confmem	Confmem;
 typedef struct FPsave	FPsave;
+typedef struct PFPU	PFPU;
 typedef struct L1	L1;
 typedef struct Label	Label;
 typedef struct Lock	Lock;
 typedef struct KMap	KMap;
 typedef struct MMMU	MMMU;
 typedef struct Mach	Mach;
-typedef struct Notsave	Notsave;
 typedef struct Page	Page;
 typedef struct Proc	Proc;
 typedef struct PMMU	PMMU;
@@ -43,9 +43,12 @@ struct FPsave
 	uchar	regs[256];
 };
 
-/*
- * FPsave.status
- */
+struct PFPU
+{
+	int	fpstate;
+	FPsave	fpsave[1];
+};
+
 enum
 {
 	FPinit,
@@ -80,13 +83,6 @@ struct Conf
 };
 
 /*
- *  things saved in the Proc structure during a notify
- */
-struct Notsave {
-	int	emptiness;
-};
-
-/*
  *  MMU stuff in proc
  */
 #define NCOLOR	1
diff --git a/sys/src/9/zynq/main.c b/sys/src/9/zynq/main.c
index c281a5330..9056f57d8 100644
--- a/sys/src/9/zynq/main.c
+++ b/sys/src/9/zynq/main.c
@@ -50,10 +50,10 @@ procfork(Proc *p)
 	s = splhi();
 	switch(up->fpstate & ~FPillegal){
 	case FPactive:
-		fpsave(&up->fpsave);
+		fpsave(up->fpsave);
 		up->fpstate = FPinactive;
 	case FPinactive:
-		p->fpsave = up->fpsave;
+		memmove(p->fpsave, up->fpsave, sizeof(FPsave));
 		p->fpstate = FPinactive;
 	}
 	splx(s);
diff --git a/sys/src/9/zynq/trap.c b/sys/src/9/zynq/trap.c
index 510860b63..cb35b4a9d 100644
--- a/sys/src/9/zynq/trap.c
+++ b/sys/src/9/zynq/trap.c
@@ -134,7 +134,7 @@ mathtrap(Ureg *, ulong)
 		break;
 	case FPinactive:
 		s = splhi();
-		fprestore(&up->fpsave);
+		fprestore(up->fpsave);
 		up->fpstate = FPactive;
 		splx(s);
 		break;
@@ -304,7 +304,7 @@ notify(Ureg *ureg)
 		return 0;
 
 	if(up->fpstate == FPactive){
-		fpsave(&up->fpsave);
+		fpsave(up->fpsave);
 		up->fpstate = FPinactive;
 	}
 	up->fpstate |= FPillegal;
@@ -517,7 +517,7 @@ procsave(Proc *p)
 		if(p->state == Moribund)
 			fpclear();
 		else
-			fpsave(&p->fpsave);
+			fpsave(p->fpsave);
 		p->fpstate = FPinactive;
 	}
 	cycles(&t);